Neural video representations
Systems for compact video encoding, fast playback, and simulation-aware memory.
Research
Representation, simulation, and physical-world modeling.
Our work sits between scientific computing and machine learning: neural representations for rich fields, learned operators for evolving systems, and tools that help AI reason about physical structure rather than surface appearance alone.
Operator learning for dynamical systems
Models for fluid transport, Rayleigh–Bénard convection, and long-horizon evolution.
WebGPU-native simulation demos
Interactive research artifacts that make physical models legible on the web.